Naser 2026's ten-model citation audit finds the same popularity-driven citation-selection bias across every major LLM vendor

chatgptllmcitation-metricsmatthew-effectbibliometricsreplicationcross-model-audit

M.Z. Naser (Clemson University) audited 69,557 citation instances generated by ten commercially deployed LLMs (GPT-5-mini, GPT-5-nano, GPT-4o-mini, Claude haiku-3.5, Claude haiku-4.5, Llama4-scout, Llama4-maverick, DeepSeek-v3.1, Kimi-k2.5, Mistral-small-3) across four academic domains, verifying references against CrossRef, OpenAlex, and Semantic Scholar. The paper cites Petiška's 2023 paper directly as prior work. Its own finding: "median cited-by counts for confirmed references range from 359 (kimi-k2.5) to 1,132 (GPT-5-nano), while field-level medians for the four domains in our study fall between 50 and 100 citations. This implies that LLMs do not seem to sample uniformly from their training distributions, but instead they preferentially retrieve highly cited works." The bias was strongest in the two most accurate (lowest-hallucination) models tested, GPT-5-mini and GPT-5-nano, at p < 10⁻⁴⁶.

Together with Algaba et al.'s peer-reviewed replication, this extends the Matthew effect finding from a single 2023 single-model study to the current generation of frontier models from every major vendor (OpenAI, Anthropic, Meta, DeepSeek, Moonshot AI, Mistral).

Caveat carried forward: Naser 2026 is, like Petiška's own paper, a single-author preprint not yet independently peer-reviewed at the time of this note. It is a second, separate unrefereed primary — it does not itself discharge the single-source concentration cap on Petiška's preprint, and its own findings should be read with the same caution Petiška's carry. (This is the first claim-note in the vault resting on this specific preprint; well under the sources.md three-note cap.)